Biologically driven cut-off definition of lymphocyte ratios in metastatic breast cancer and association with exosomal subpopulations and prognosis

High neutrophil to lymphocyte ratio (NLR) and monocyte to lymphocyte ratio (MLR) are respectively associated with systemic inflammation and immune suppression and have been associated with a poor outcome. Plasmatic exosomes are extracellular vesicles involved in the intercellular communication system that can exert an immunosuppressive function. Aim of this study was to investigate the interplay between the immune system and circulating exosomes in metastatic breast cancer (MBC). A threshold capable to classify patients according to MLR, NLR and PLR, was computed through a receiving operator curve analysis after propensity score matching with a series of female blood donors. Exosomes were isolated from plasma by ExoQuick solution and characterized by flow-cytometry. NLR, MLR, PLR and exosomal subpopulations potentially involved in the pre-metastatic niche were significantly different in MBC patients with respect to controls. MLR was significantly associated with number of sites at the onset of metastatic disease, while high levels of MLR and NLR were found to be associated with poor prognosis. Furthermore, exosomal subpopulations varied according to NLR, MLR, PLR and both were associated with different breast cancer subtypes and sites of distant involvement. This study highlights the nuanced role of immunity in MBC spread, progression and outcome. Moreover, they suggest potential interaction mechanisms between immunity, MBC and the metastatic niche.

Statistical analysis. Clinico-pathological characteristics of the MBC_main cohort were summarized through descriptive analysis. Categorical variables were described through frequency distribution, whereas continuous variables were reported through median and interquartile range. MLR, NLR, PLR and exosomal subpopulations distributions were tested for normality by Shapiro-Wilk test and their association with metastatic disease and clinico-pathological features was explored by Wilcoxon rank-sum test or Kruskal-Wallis test, as statistically appropriate.
To identify a threshold capable to classify patients according to MLR, NLR and PLR, a receiving operator curve (ROC) analysis was performed, after propensity score matching (PSM) with a series of 1228 female blood donors. Propensity score was computed by taking into consideration age. The matching approach was 1:1 nearest neighbor with caliber of 20% 29,30 . The optimal threshold was calculated through empirical cut point estimation with Liu method 31 .
Overall survival (OS) was defined as the time between treatment start and death from any cause. Prognostic factors in terms of OS were tested both in uni-and multivariate models by Cox regression with 95% confidence interval (95% C.I.). Differences in survival were tested by log-rank test and represented by Kaplan-Meier estimator plot.
Variations in MLR, NLR and PLR between first and second line were analyzed through Wilcoxon signed-rank test.
Statistical analysis was performed using STATA (StataCorp. Blood sample analysis. NLR and PLR were defined as the absolute neutrophil count divided by the absolute lymphocyte count, and the absolute platelet count divided by the absolute lymphocyte count, respectively; MLR was the result of the ratio between the absolute monocyte count and absolute lymphocyte count. These parameters were analyzed from peripheral blood cell count by DxH 800 hematology analyzer (Beckman Coulter). Full blood count data were eligible for analysis if performed within 2 months before the start of first line and 3 days for second-line. exosome precipitation and characterization. Exosome precipitation. Exosome assessment was performed before the beginning of a new therapeutic line. Exosome were precipitated from 1 ml of "platelet-free plasma" by ExoQuick Exosome Precipitation Solution (SBI System Biosciences) according to manufacturer's protocol 25 . Briefly, fresh serum was centrifuged 30 min at 3000 g at 4 °C. To 250 µl of serum were added 63 µl of ExoQuick and incubated at 4 °C 30 min. After this period, the sample was centrifuged 45 min at 1000 g and the supernatant was eliminated. Again, the pellet was centrifuged 5 min at 1000 g to remove all traces of fluid and exosome pellet was resuspended in 200 μl of Dulbecco's Phosphate Buffered Saline (PBS). "Platelet-free plasma" samples were obtained from ethylenediamine tetraacetic acid (EDTA)-added peripheral blood samples by two sequential centrifugations (2600 rpm for 15 minutes at room temperature (RT) followed by the centrifugation of the supernatant at 13000 g for 15 minutes at 4 °C). Before exosome precipitation, "platelet-free plasma" was pre-treated with 5U/ml Thrombin (SBI System Biosciences) following manufacturer instructions.
Exosome purification by sucrose linear density gradient. In order to demonstrate the presence of exosomes in the ExoQuick enriched fractions, Fresh serum was centrifuged 30 min at 3000 g at 4 °C. To 500 µl of serum were added 126 µl of ExoQuick (System Biosciences) and incubated at 4 °C 30 min as reported in the manufacturer's protocol. After this period, the sample was centrifuged 45 min at 1000 g and the supernatant was eliminated. Again, the pellet was centrifuged 5 min at 1000 g to remove all traces of fluid and exosome pellet was resuspended in 100 μl of D-PBS. Exosomes were then isolated by floatation in linear sucrose gradient. Briefly, exosomes isolated by Exoquick (100 μl) were resuspended in 400 µl of 20 mM Hepes pH 7.4 containing 2.50 M sucrose. The samples were transferred into polyallomer centrifuge tube, then carefully overlaid (by peristaltic pump) with continuous sucrose gradient (from 2.00 M sucrose in 20 mM Hepes pH 7.4 to 0.25 M sucrose in 20 mM Hepes pH 7.4). Centrifugation was performed for 16 h at 40000 rpm, 4 °C (Optima Max ultracentrifuge and MLS-50 rotor; thinwall polyallomer tube). After ultracentrifugation, ten fractions of 0.5 ml were recovered from top (fraction 1) to bottom (fraction 10). Each gradient fraction was subjected to ultracentrifugation to eliminate sucrose and concentrate exosomes. Fractions (0.5 ml) were diluted with 2.5 ml of 20 mM HEPES, pH 7.4 and centrifuged in 3 ml tubes 1 h at 50000 rpm, 4 °C (Optima Max ultracentrifuge and TLA-100.3 rotor; thickwall polycarbonate tubes). Exosome quantification was performed by Bradford's assay after lysis of the samples in RIPA lysis buffer (NaCl 150 mM, 1X NP40, 0.1% SDS, 1% sodium deoxycholate, 25 mM Tris HCl pH 7.6, all of Sigma-Aldrich) in the presence of protease inhibitors (ThermoScientific).
Atomic force microscopy (AFM) analyses. AFM analysis was carried out adsorbing for 30 minutes 100 μl of exosomes enriched by ExoQuick and diluted in PBS on freshly cleaved 11 × 11 mm mica sheets (Agar Scientific). Imaging was performed in liquid (PBS) on a MFP-3D Stand Alone (Oxford Instruments GmbH, Wiesbaden, Germany) in dynamic mode with silicon probes (Force constant 0.5-1 N/m, radius of curvature <10 nm, NSC36 Mikromasch, Sofia, Bulgaria). Topographic height images were acquired at 256×256 and 512 × 512 pixels with a scan rate of 1-2 Hz. Image processing and data analysis has been performed using Igor Pro and Gwyddion softwares.
Nanoparticle tracking analysis (NTA) analyses. Assessment of exosomes by NTA was performed on a NanoSight LM10 system (Malvern) by analyzing ∼500 μl of ExoQuick-enriched exosome preparations properly diluted in PBS (∼10 3 -10 4 times). Individual videos of 60 seconds for each sample were acquired using the maximum camera gain and analyzed by the NanoSight particle tracking software to determine particles size and density.
Western blot analysis. Samples were resuspended in sample buffer 1X and subjected to 10% Sodium Dodecyl Sulphate -PolyAcrylamide Gel Electrophoresis (SDS-PAGE) under reducing or non-reducing conditions, respectively. Proteins were transferred to Polyvinylidene difluoride (PVDF) membrane and reacted with primary antibodies, overnight at 4 °C, at the following dilutions: 1:500 for mouse monoclonal anti-CD9, 1:500 for mouse monoclonal anti-CD63. After being washed, the membranes were incubated with secondary anti-mouse IgG and developed by Enhanced ChemiLuminescence (ECL).
Dynamic light scattering (DLS). Exosomes from pooled CD9-positive fractions 3-5 were subjected to DLS. DLS is a technique for measuring the size and size distribution of molecules and particles dispersed or dissolved in a liquid. The Brownian motion of particles or molecules in suspension causes laser light scattering at different intensities. Analysis of these intensity fluctuations yields the velocity of the Brownian motion and hence the particle size using the Stokes-Einstein relationship.
Scanning electron microscope (SEM) analysis. CD9 positive fractions (Fractions 3-6) were thawed from a minus 80 °C freezer and prepared for SEM imaging. Both samples were diluted properly (at least 100 times) in PBS (pH 7.4) and volume of 10 μl of each dilution were deposited on a pure, thin glass substrate. The plates with fixed exosomes were stored in 4 °C temperature for gentle drying. So-prepared exosomes underwent immediate gold/ palladium (80:20, 60 seconds) sputtering for scanning electron microscopy visualization. Structure of exosomes was analysed by scanning electron microscopy SEM (Jeol, JSM 7001 F TTLS). The etched surfaces were coated for 60 seconds with gold/palladium (80:20) using sputter coater/turbo evaporator (Quorum Technologies Q150T ES) in order to provide an electrically conductive thin film to reduce thermal damage and charging of the samples. The SEM micrographs were acquired by applying the accelerating voltage of 15 kV and SEI secondary electron methodology.
Cytofluorimetric analysis of exosomes. Because of the resolution limit of a flow-cytometer, exosomes were bound to 4-μm aldehyde/sulfate latex beads (Molecular Probes, Invitrogen).
Briefly, latex beads were firstly functionalized with anti-CD63 antibodies by mixing and incubating overnight at 4 °C equal amounts of 0.5 mg/ml anti-CD63 antibodies (BD Biosciences) and 20 mg/ml latex beads. After blocking of the reaction by 0.1% glycine (Sigma-Aldrich), latex beads are re-suspended in PBS/0.1% BSA at the final concentration of 300,000 beads/μl. Exosomes are bound to latex beads by incubating 10 μg of exosomal proteins, resuspended in 100 μl of 1x PBS, with 2 μl of CD63-coated beads, corresponding to 200000 CD63-coated beads, for 15 at room temperature. Reactions were stopped by 0.1% glycine in PBS and, after washes with 0.05% PBS/Tween, incubated with 2 μl of anti-Fc receptor (Miltenyi Biotec).

Results
LRs are significantly different with respect to healthy controls. This retrospective study analyzed a consecutive cohort of 396 patients with a diagnosis of MBC between January 2006 and 2017 (MBC_main patients) and 1128 blood donors. Among the analyzed patients, 42 were younger than 45 years, 186 were aged between 45 and 65 years, and 168 were older than 65 years. Median follow-up was 53 months, and the median overall survival was 31 months. Survival was 78.56% at 12 months, 55.98% at 24 months, and 27.49% at 60 months. Patients' clinico-pathological characteristics are described in Table 1.
NLR, MLR and PLR were significantly different in MBC_main patients with respect to blood donors (p = 0.0001 for MLR, NLR and PLR). On the other hand, no significant differences were observed according to the type of metastatic onset (i.e. de novo versus relapsed).
In order to calculate a threshold for NLR, MLR, PLR capable to stratify patients' population, a ROC analysis was performed on a propensity score matched pooled cohort comprising both blood donors and MBC patients. The optimal cut-off identified for MLR was 0.28 (AUC = 0.6447), 2 for NLR (AUC = 0.7439), and 148.38 (AUC = 0.6766) for PLR. Patients with LRs equal or above these thresholds were defined as MLR high , NLR high and PLR high , respectively.

Scientific RepoRtS |
(2020) 10:7010 | https://doi.org/10.1038/s41598-020-63291-2 www.nature.com/scientificreports www.nature.com/scientificreports/ Exosomal subpopulations are significantly different with respect to healthy controls. Plasma exosomes were precipitated by using ExoQuick, a commercial kit based on the use of polyethylene glycol (PEG). Atomic Force Microscopy ( Fig. 1A-C) and Nanoparticle Tracking Analysis (NTA) by Nanosight (Fig. 1D) confirmed the presence, in precipitated preparations, of vesicles with a diameter compatible with that of exosomes (10 th , 50 th and 90 th percentiles: 67.4 nm, 99.1 nm and 155.8 nm, respectively; n = 10), with an average concentration of the order of 10 12 particles per ml of plasma. To verify that the isolated particles were enriched in exosomal markers we first analyzed, by western blot, the expression of markers such as CD9, Alix and enolase. Since Exoquick is able to precipitate also the proteins of the plasma, diluting the amount of exosomal proteins, Exoquick-enriched preparations were further subjected to floatation on linear sucrose density gradient (2-0.25 M; Fig. 1E,F). Of the 10 fractions collected it was shown that fractions 3 to 5 of the gradient were enriched in CD9 (Fig. 1G). The same fractions analyzed by SEM and DLS showed the presence of particles whose morphology and size were compatible with exosomal particles (Fig. 1H,I).
Once demonstrated that ExoQuick preparations were indeed enriched in exosomes, we proceeded with a more extensive phenotypic characterization of exosomes adopting a cytofluorimetric protocol meant to detect  www.nature.com/scientificreports www.nature.com/scientificreports/ and characterize exosomal subpopulations on a multiparametric basis. Since exosomes are smaller than the cytofluorimeter resolution limit, 4μm latex beads functionalized with anti-CD63 antibodies were used to capture extracellular vesicles. The presence of these latter was subsequently confirmed by the positivity for CD9 and CD63, and CD9/CD63 positive subpopulations were finally analyzed for the expression of different markers, i.e. the fraction of exosomes expressing markers of epithelial origin (EpCAM, E-Cadherin, EGFR and HER2), markers of exosomes possibly involved in the metastatic process (HGF-R, CD44, CD49d, N-Cadherin, CXCR4), immune-related markers (CD45, CD13, CD11a, CD104), and platelet-derived (CD61) and endothelial-derived (CD31) exosomes ( Fig. 2A-C). Comparing 20 healthy donors with 20 MBC patients, differences in the relative composition of plasma exosomes were detected, as displayed in Fig. 2D. Specifically, the fraction of exosomes expressing epithelial markers, such as EpCAM, HER2 and E-Cadherin, was very low in donors and increased, although representing a small fraction of exosomes, in MBC patients, reaching a statistical significance for EpCAM and E-Cadherin. The expression of markers potentially involved in the pre-metastatic niche, such as HGF-R, CD44 and CXCR4 was significantly higher in patients than in controls. These markers were expressed in a fraction of exosomes that reached 10% in patients. No significant differences in the fraction of N-Cadherin, CD61 and markers linked to the immune system were detected among MBC patients and donors. For these reasons, the MBC_exo cohort was assessed for exosomes expressing epithelial and metastatic markers.

MBC-derived exosomes inhibit T cell proliferation and are preferentially internalized by circulating monocytes.
We have previously shown that in glioblastoma patients, plasma-derived exosomes were able to reduce the proliferation of T lymphocyte and that this effect was mediated by CD14 + monocytes 28 . To establish whether also exosomes derived from the plasma of MBC patients were able to reduce the proliferation of T lymphocyte, CFSE-labeled PBMCs, isolated from healthy donors, were pretreated for 24 hours without or with plasma-derived exosomes and stimulated or not for 72 hours with anti-CD3 and anti-CD28. As shown in Fig. 3, the proliferation of T-lymphocytes was abolished by MBC-derived exosomes at both the concentration used (1:2 and 1:100). Conversely, exosomes isolated from the plasma of healthy donors did not affect T cell proliferation (Fig. 3). Additionally, by incubating PBMC with DiD labeled exosomes for 5 or 24 hours, it has been shown that, at both time points, 99.9% of CD14 + monocytes resulted to be DiD-positive, while only a minor fraction of CD3 + T lymphocytes internalized exosomes, independently form the time point (Fig. 4).
Therefore, like in glioblastoma patients, exosomes derived from MBC patients were mainly up-taken by monocytes and suppressed T-cell proliferation, thus posing the rationale for an overall analysis focused on both exosomes and NLR/MLR. Exosomal subpopulations vary according to LRs and are both associated with different breast cancer subtypes. The stratification of MBC_main according to BC subtypes, highlighted a significantly higher NLR among patients affected by non-luminal HER2-positive and triple negative disease (p = 0.0019). No statistically significant differences were highlighted for MLR or PLR (Fig. S1). In MBC_exo, MLR high and PLR high patients with HER2 BC had a significant higher fraction of CD49d pos exosomes (P = 0.0086, P = 0.0166 respectively) and a trend for CXCR4 pos . Patients in the NLR low subgroup showed higher fraction of EPCAM pos (P = 0.0344) when affected by HER2 positive disease and CXCR4 pos for ER positive disease (P = 0.0344).
MLR high patients of the MBC_exo cohort had a higher fraction of HGFR pos exosomes (P = 0.0055), while PLR high patients had a higher fraction of HER2 pos and EpCAM pos exosomes (P = 0.04). Lymphocyte count higher than 1.8 × 10 3 μL was associated with low HER2 pos and high EpCAM pos levels (P = 0.0082 and P = 0.0193).
LRs are associated with tumor burden and to different exosomal subpopulations according to sites of distant involvement. When considering the association between MLR, NLR, PLR and number of metastatic sites in the MBC_main cohort, only MLR was significantly associated with number of sites at the onset of metastatic disease (p = 0.0096); on the other hand, NLR was associated with an increase in number of sites (p = 0.036). PLR was associated neither with tumor burden, nor with their increase. Notably, no direct associations were observed between LRs and specific metastatic sites.
Consistently with the hypothesis that exosomes could influence the metastatic niche, we observed that patients in with visceral metastases in the MBC_exo cohort had a higher fraction of vascular endothelial growth factor receptor 2 pos (VEGFR-2) (P = 0.0323) and marginally of E-caderin pos exosomes, while lung metastases were marginally associated with higher expression of VEGFR-2 pos exosomes.
For exploratory purpose, MBC_exo was stratified according to lymphocyte ratios in order to highlight differences according to the immune and inflammatory status. Patients with bone localizations had a higher proportion of EGFR pos exosomes in the MLR low , NLR low and PLR low subgroups (P = 0.035, P = 0.01 and P = 0.016 respectively). Notably, patients in the NLR high and PLR high subgroup had high levels of VEGFR pos exosomes when high tumor burden (i.e ≥ 2 metastatic sites) or visceral localizations were detected (NLR high P = 0.0158 and P = 0.0052; PLR high P = 0.039 and 0.027, respectively). Furthermore, patients in the NLR high subgroup had a lower fraction of circulating CD49d exosomes if lung involvement was detected (P = 0.036) and higher VEGFR2 in case of liver localizations (P = 0.037) (Fig. 5).

LRs impact on prognosis and its magnitude varies across Bc subtypes.
On univariate analysis, MLR high and NLR high patients in the MBC_psm cohort experienced poor prognosis (HR 1.77, 95% CI: 1.24-2.54, p = 0.002; HR 2.09, 95% CI: 1.40-3.12, p < 0.001, respectively). Similar results were observed also in the MBC_ main population (data not shown). On the other hand, no significant prognostic impact was observed for PLR (Table 2) (Fig. 6).
Based on these data, MLR and NLR were combined in a score in order to explore the interplay of inflammation and immunocompetence on prognosis. The score ranged from 0 (both MLR and NLR under the cut-off) www.nature.com/scientificreports www.nature.com/scientificreports/ through 1 (only one between MLR and NLR above the cut-off) to 2 (both ratios above the cut-off). On univariate analysis, the score was capable to stratify the MBC_psm population in 3 groups according to prognosis (OS) (score 1 vs 0 HR 2.35 95% CI: 1.38-4.02, p = 0.002, score 2 vs 0 HR 2.54 95% CI: 1.59-4.07, p < 0.001) (Fig. S2). These results were confirmed in the MBC_main population (data not shown).
Multivariate analysis of the MBC_psm subgroup, showed that only MLR (HR 1.85 95% CI = 1.21-2.83, p = 0.005) and NLR-MLR score 2 vs 0 (HR 1.90 95% CI 1.11-3.26, p = 0.020) retained a significant prognostic impact on OS (Table 3). When applying the models to MBC_main, neither MLR nor NLR nor the NLR-MLR score were independently associated with prognosis (Table 3). www.nature.com/scientificreports www.nature.com/scientificreports/ A subgroup analysis was then performed on the MBC_main population to investigate the different impact of lymphocytes ratios according to clinico-pathological features. Test for heterogeneity highlighted a significant interaction according to age for MLR (P for interaction = 0.028) and a trend for NLR (P for interaction = 0.088) (Fig. 7).  Table 3).
The prognostic impact of variation in MLR, NLR and PLR from the first to the second line of therapy was also investigated. Patients of MBC_main with decreased or stable ratios, experienced a better outcome for all the three parameters taken into consideration (i.e. MLR p = 0.028, NLR p = 0.034 and PLR p = 0.003) (Fig. 8). www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
The present study examined a cohort of 396 patients affected by MBC and investigated the association between NLR, MLR, PLR and exosomal subpopulations. Moreover, the role of LRs on prognosis was explored and its magnitude was related to BC subtypes. LRs have been intensively studied in the last few years in several cancer types, including breast cancer, and the results observed in our study are consistent with those found in literature.
Notably, previous studies based their cut-offs according to literature or on the analyzed population's prognosis or division in quartiles. The present study adopted an alternative approach using a pooled database composed by propensity score matched blood donors and MBC patients. This strategy highlighted significantly higher values of NLR, MLR and PLR in MBC patients compared to healthy subjects and enabled us to calculate a cut-off on a biological basis and not related to the intrinsic characteristics of the analyzed population. The obtained cut-offs www.nature.com/scientificreports www.nature.com/scientificreports/ for NLR and PLR are slightly lower with respect to those found in literature (NLR: 1.9; literature range: 2.0 to 5.0, PLR: 150; literature range: 160-300 7,9,32 ). Few studies, to date, have investigated the role of monocytes-based parameters in MBC.
This study, moreover, suggests some important aspects linked to the interaction between cancer cells, immunity and the tumor micro-environment. In fact, as previously shown in glioblastoma patients 28 , exosomes isolated from the plasma of MBC patients exerted an immunosuppressive function on T-lymphocytes and were preferentially internalized by monocytes. In a murine model it has been shown that breast cancer-derived exosomes alter macrophage polarization possibly inducing in monocytes a skewing to a cancer-promoting phenotype 33 . Intriguingly, exosomal subpopulations linked to immunity (i.e. CD49d and CXCR4) vary significantly according to HER2, ER status and MLR, NLR and PLR. This result reinforces the notion of a possible connection between specific exosomal subpopulations and immune system. Beside their role in immunity, exosomes are considered to play a significant role in the metastatic spread of a primary tumor, both by participating in oncogenic reprogramming of malignant cells and by contributing to formation of a pre-metastatic niche, thus favoring exosome-mediated organ-specific metastasis 34,35 . Accordingly, results concerning the association of distinct exosomal subpopulations with distant cancer dissemination underline the strict interplay between the immune system and the metastatic niche. The observed reduced frequency of CD49d exosomes, in NLR high patients with lung localizations, could be a consequence of the clustering activity of this integrin during interstitial lung inflammation and secondary to the pivotal role played by neutrophils in the formation of the lung pre-metastatic niche 36,37 . Indeed, in vitro studies have shown that CD49d can bind to fibronectin and facilitate C5a-induced neutrophil migration across lung fibroblast monolayers 38,39 . Neutrophils consequently alter the endothelial permeability and the extracellular matrix through lysine oxidase induced collagen cross-linking and matrix metalloproteinases collagen-disruption, potentially favoring circulating tumor cells seeding 36,37 .
Patients with visceral metastasis and high tumor burden (i.e ≥2 metastatic sites) showed increased levels of VEGFR2 pos exosomes, suggesting the importance of vascular integrity and neoangiogenesis for the metastatic spread 40 . Interestingly, patients with bone metastasis showed a higher fraction of EGFR pos exosomes in the MLR low , NLR low and PLR low subgroups, suggesting a potentially different homing mechanism with respect to visceral involvement. Indeed, it has been pre-clinically demonstrated, that EGFR ligands (e.g. EGF, TGF-β, and amphiregulin) are able to directly stimulate osteolysis when expressed in the bone microenvironment 41,42 . Furthermore, EGF and TGF-β mediated osteoclastogenesis is accompanied by a sustained production of RANKL by bone stromal cells, potentially favoring breast cancer's bone tropism 43,44 . Notably, patients with visceral  www.nature.com/scientificreports www.nature.com/scientificreports/ localizations were found to have higher EPCAM pos and ECAD pos , exosomes consistent with the evidence that surface adhesion molecules displayed on tumor-derived exosomes may influence the metastatization pattern 36,45,46 .
Moreover, elevated MLR and NLR were associated with a worse prognosis, and their combination was capable to further stratify the population in three different prognostic groups 32 . PLR, on the contrary, was found not significantly associated with OS, in accordance to a previous large meta-analysis, but in contrast with the work by Zhu et al., which reported a negative prognostic impact on breast cancer patients with a high PLR in terms of OS 9,47 .
The obtained cut-offs are consistently significant when analyzed in the matched population, but some exceptions were found when applied to the general population.
One major reason could be represented by fact that donors older than 65 years old were only 7, and since lymphocyte ratios are deeply affected by the subject's age, this could have calibrated the thresholds on a younger population of patients. This hypothesis is consistent with the subgroup analysis that highlighted a significant P for interaction for MLR. Furthermore, the subgroup of patients above 65 years is more prone to develop comorbidities and competing risk factors that could hinder a lean evaluation of survival.
Even though several pieces of evidence have shown the pivotal role played by the immune system in triple negative and HER2 positive BC, data concerning lymphocyte ratios in this setting are still widely debated 1,48 . www.nature.com/scientificreports www.nature.com/scientificreports/ A recent meta-analysis conducted by Xing et al., showed that NLR was found to be associated with OS in HER2 positive and TNBC subtypes, whereas no association was found in luminal A and luminal B tumors 49 . Ethier and colleagues, however, found that the negative prognostic effect of NLR on OS was consistent in all clinico-pathological groups 32 .
This finding has been observed in our cohort when NLR levels were compared between molecular subgroups. Moreover, such differences could explain why MLR retained its significance when the multivariate analysis was restricted to triple negative and HER2-positive breast cancer patients (HR 1.92, 95% CI: 1.15-3.22, p = 0.013).
Jia et al., showed that NLR is stronger than MLR in predicting OS in all BC subtypes, as opposed to our results. However, both NLR and MLR, consistently failed to retain their significance among patients affected by luminal-like breast cancer 50 . Conversely Ji et al., found LMR to be associated with survival in luminal BC patients 13 . The weak prognostic impact of NLR observed in our cohort could be due to the high proportion of patients with luminal-like disease. In these cases, NLR was remarkably lower, with values often close to the cut-off.
Since the interaction between immunity and tumor cells is dynamic, analyzing of the impact of lymphocyte ratio's fluctuations might be crucial.
Previous studies assessed NLR, MLR and PLR before surgery or before the beginning of systemic treatments, limiting the evaluation of their prognostic role only at a pre-treatment phase. However, chemotherapy and especially tumor burden can influence chronic inflammation and therefore can have an impact of the lymphocyte www.nature.com/scientificreports www.nature.com/scientificreports/ ratios. Conversely, disease progression or changes in the molecular phenotype of cancer, can be reflected by an alteration in NLR, MLR and PLR.
Our data showed that variation of these ratios is associated with worse prognosis, in particular when an increase is detected. When considering tumor burden, only NLR was found to be associated with an increase in metastatic sites (p = 0.04). However, it is important to notice that the number of metastatic sites might not reflect the real tumor burden and that progression may also occur without an increase in the number of sites. Therefore, further evaluations are required in order to understand the relationship between MLR, NLR, PLR, tumor burden and progression.
Although exploratory, the present study gives distinct insights that could have several immediate and future clinical applications. These results offer biologically driven LR cut-offs that could be used not only for patients' stratification and clinical trial design but also for clinical monitoring, since longitudinal variations in LRs are informative of the disease evolution. Moreover, as immunotherapy is gradually gaining momentum also in TNBC, cost-effective immunity-driven biomarkers could have an interesting role as observed in other cancer types [51][52][53] .
Notably, exosomes are capable to influence the immune system by acting on monocyte maturation, therefore posing some intriguing perspectives on the integration of LRs with the exosomal characterization 28 . If on one hand these results pose the basis for a potential monitoring tool capable to highlight sites that could be more prone to distant involvement, on the other, the study confirms the interaction between the metastatic niche and the immune system and offers an interesting rationale for targeting such interactions for future treatment strategies.

conclusions
The present study was able to demonstrate that elevated monocyte-to-lymphocyte and neutrophil-to-lymphocyte ratios are significantly associated with cancer, as they present lower values in healthy subjects. These ratios significantly vary among different breast cancer subtypes, reflecting the different involvement of the immune system in HER2-positive and triple negative disease. Furthermore, monocyte-to-lymphocyte and neutrophil-to-lymphocyte ratios are confirmed as prognostic factors and the proposed score suggested their possible combination in order to stratify patients younger than 65 years old in three different risk subgroups that consider aspects linked both to systemic inflammation and immunodepression.
Interestingly, the reported findings suggested an important interplay between immunity, specific circulating exosomal subpopulations and metastatic sites, suggesting the possibility to infer information on the organ-specific metastatic process.
Albeit the purpose of the study was exploratory, these data suggest that further investigations should be considered on this topic to better dissect immunoediting and homing mechanisms of circulating tumor cells.